Module chunker 

Semantic windowing chunker for RAG indexing.

Splits timeline entries into overlapping windows of ~1.5k tokens, suitable for vector embedding and semantic search via memex.

Vibecrafted with AI Agents by VetCoders (c)2026 VetCoders

Structs

Chunk

A single chunk ready for vector indexing.

ChunkMetadataSidecar

Structured metadata sidecar persisted alongside each chunk file.

ChunkerConfig

Configuration for the chunker.

Constants

INTENT_KEYWORDS

Functions

chunk_entries

Chunk timeline entries into semantic windows with overlap.

chunk_summary

Summary of chunking results.

classify_kind

Classify a set of timeline entries into a canonical Kind.

estimate_tokens

Estimate token count from text length.

format_chunk_text

Format entries into chunk text with metadata header.

is_decision_tag

is_outcome_tag

is_result_line

normalize_key

parse_checklist_task

truncate_signal_line

write_chunks_to_dir

Write chunks as individual .txt files to a directory.